Preliminary study for the assessment of discontinuity’s size through Machine Learning algorithms

During the last two decades there has been a huge breakthrough in the Structural Health Monitoring field, especially in the study of Acoustic Emissions (AE), to get qualitative and quantitative damage-related information. This thesis attempts to focus on the possibility of obtaining an automatic estimate of small discontinuity’s length in an aluminium plate, by analysing some impinging signals when they interfere with the defect itself. The novel aspect about this analysis is that it was conducted through “trained” classification and regression algorithms that have been able, up to some extent, to automatically classify and predict the desired responses. This means that Artificial Intelligence, in particular Machine Learning techniques, were employed and played an important role within either the identification and the predictive part of this study. Due to the SARS-CoV-2 global pandemic, and the consequent closure of the US embassies, it was not possible to obtain the Visa and go to Clarkson University to perform the experimental campaign there. Therefore, in order to collect the raw signals for the subsequent analysis, a comparison between Abaqus CAE and OnScale software was firstly enforced, and eventually the latter was chosen to perform the whole set of numerical simulations exploiting a pitch-catch configuration.